Strain HTCC1062 was cultured in AMS1 medium amended with 25 μM glycine, 10 μM methionine and 50 μM pyruvate at 16°C according to the reported protocol (Rappe et al., 2002; Tripp, 2013). AMS1 was sparged with CO₂ for 5 h followed by sparging with air for 10 h. The pH of the resulting AMS1 typically ranged from 7.5 to 7.7. Cells of strain HTCC1062 were stained with SYBR Green I (Molecular Probes, America) and counted with a Guava Technologies flow cytometer (Millipore, America). The E. coli strains DH5α, BL21(DE3) and WM3064 were grown in the Lysogeny Broth (LB) medium at 37°C. Diaminopimelic acid (0.3 mM) was added into the LB medium to culture E. coli WM3064. R. pomeroyi DSS-3 was purchased from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures and was cultured in 974 medium at 30°C according to the protocol provided¹.

Strain HTCC1062 was firstly cultured in AMS1 medium amended with 25 μM glycine, 10 μM methionine and 50 μM pyruvate. When the concentration of cells reached 2 × 10⁷ cells/ml, TMAO was added into the medium with a final concentration of 0.8 mM. The group without the addition of TMAO was set up as a control. After 0.5 or 2 h incubation, RNA was extracted from the cells using the RNeasy mini kit (Qiagen, America), and was subsequently reverse-transcribed to cDNA using Goldenstar™RT6 cDNA Synthesis Kit (TsingKe, China). The qPCR experiments were performed using a Light Cycler II 480 System (Roche, Switzerland) following the instructions of SYBR^® Premix Ex TaqTM (TaKaRa, Japan) with the following cycling conditions: 95°C for 5 min, 45 cycles of 95°C for 10 s and 60°C for 30 s. The recA gene was used as an internal reference gene.

Deletion of the tmoW gene of R. pomeroyi DSS-3 was performed by pK18mobsacB-Ery based homolog recombination (Wang et al., 2015). The upstream and downstream sequences of the tmoW gene were amplified with primer sets tmoW-UP-F/tmoW-UP-R and tmoW-Down-F/tmoW-Down-R (Supplementary Table 1 and Supplementary Figure 1). Then, the PCR fragments were inserted to the vector pK18mobsacB-Ery with HindIII/BamHI as the restriction sites to generate pK18Ery-tmoW, which was transferred into E. coli WM3064. Next, the plasmid pK18Ery-tmoW was mobilized into R. pomeroyi DSS-3 by intergeneric conjugation with E. coli WM3064. To select for colonies in which the pK18Ery-tmoW had integrated into the R. pomeroyi DSS-3 genome by a single crossover event, cells were plated on the marine 2,216 agar plates containing erythromycin (25 μg/ml). Subsequently, the resultant mutant was cultured in the marine broth 2,216 medium and plated on the marine 2,216 agar plates containing 10% (w/v) sucrose to select for colonies in which the second recombination event occurred. For complementation of the ΔtmoW mutant, the tmoXWV₁₀₆₂ gene cluster with its native promoter was amplified from the genomic DNA of HTCC1062 with primer sets tmoXWV₁₀₆₂-350Up-F/tmoXWV₁₀₆₂-Down-R (Supplementary Table 1). The PCR fragments were digested with BamHI and EcoRI, and then inserted into the vector pHG101 to generate pHG101-tmoXWV₁₀₆₂. This plasmid was then transformed into E. coli WM3064, and mobilized into the ΔtmoW mutant of R. pomeroyi DSS-3 by conjugation.

The full-length tmoX₁₀₆₂ gene was amplified from the genomic DNA of HTCC1062 by PCR using FastPfu DNA polymerase (TransGen Biotech, China), and was subcloned into the NdeI/XhoI restriction sites of the pET22b (Novagen, America) vector with a C-terminal His-tag. All of the point mutations in tmoX₁₀₆₂ were performed with the QuikChange^® mutagenesis kit II (Agilent, America). The wild-type (WT) TmoX₁₀₆₂ protein and all of the mutants were expressed in E. coli strain BL21(DE3). The recombinant E. coli strains were cultured at 37°C in LB medium to an OD₆₀₀ of 0.8–1.0 and then incubated at 16°C for 16 h with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) as an inducer for recombinant protein expression. The recombinant proteins were purified first with Ni-affinity column (GE Healthcare, America), and then with gel filtration on a Superdex-75 column (GE Healthcare, America) eluted with the buffer containing 10 mM Tris–HCl (pH 8.0) and 100 mM NaCl. Approximately 2 mg recombinant TmoX₁₀₆₂ protein was obtained from 1 liter of culture.

Isothermal titration calorimetry (ITC) measurements were performed using a PEAQ-ITC system (Malvern, Britain). The sample cell was loaded with 250 μl of protein sample (30 μM), and the reference cell contained distilled water. The syringe was filled with 70 μl of TMAO (200 μM). The proteins and TMAO were kept in the same buffer containing 10 mM Tris–HCl (pH 8.0) and 100 mM NaCl. Titrations were carried out by adding 0.4 μl of TMAO for the first injection and 1.5 μl for the following 12 injections, with stirring at 750 rpm/min.

Wild-type TmoX₁₀₆₂ and all of the mutants were subjected to circular-dichroism (CD) spectroscopic assays at 20°C on a J-1500 spectropolarimeter (Jasco, Japan). CD spectra of the samples at a final concentration of approximately 10 μM were collected from 250 nm to 200 nm at a scan speed of 200 nm/min with a bandwidth of 1 nm. All of the samples were in a buffer containing 10 mM Tris–HCl (pH 8.0) and 100 mM NaCl. To determine the T_m of TmoX₁₀₆₂, the temperature was raised from 20 to 80°C in 1 h.

To investigate whether strain HTCC1062 can grow with TMAO as a nitrogen source, we replaced (NH₄)₂SO₄ in AMS1 medium by TMAO. Methionine, which is usually used as the reduced sulfur, was also replaced by dimethylsulfoniopropionate (DMSP) to avoid the possible interference of its nitrogen atom. As shown in Figure 2A, strain HTCC1062 showed noticeable growth in the medium with TMAO as the sole nitrogen source, although its growth on TMAO was much weaker compared to that on (NH₄)₂SO₄. This result suggests that strain HTCC1062 should contain TMAO transporter and enzymes involved in TMAO metabolism.

In R. pomeroyi DSS-3, TMAO can be transported into the cell through a TMAO specific transporter TmoXWV (Lidbury et al., 2014), and is then utilized as a nitrogen and energy source with the catalysis of several enzymes, including Tdm, DmmDABC and GmaS (Lidbury et al., 2015). Genomic analysis suggests that strain HTCC1062 possesses tmoXWV, tdm and gmaS homologs (tmoXWV₁₀₆₂, tdm₁₀₆₂ and gmaS₁₀₆₂, respectively). However, no dmmDABC homolog was identified from the genome of strain HTCC1062 (Lidbury et al., 2017; Figure 1B). RT-qPCR analysis showed that the transcriptions of tmoX₁₀₆₂, tdm₁₀₆₂ and gmaS₁₀₆₂ were all up-regulated by TMAO (Figure 2B), suggesting that these genes may be functional in TMAO import and metabolism. Thus, the SAR11 bacterial strain HTCC1062 may import and metabolize TMAO via a pathway generally similar to that of the MRC bacterial strain DSS-3, except that strain HTCC1062 may recruit an isoenzyme of DmmDABC to convert DMA to MMA. Next, we characterized TmoXWV₁₀₆₂ of strain HTCC1062 to investigate how SAR11 bacteria import TMAO in this study.

It has been reported that the tmoW-deleted mutation in R. pomeroyi DSS-3 disables its capacity to grow with TMAO as the sole nitrogen source (Lidbury et al., 2014). TmoXWV₁₀₆₂ of strain HTCC1062 shares ∼41% sequence identity to the functional TmoXWV of R. pomeryi DSS-3. Because currently genetic manipulation cannot be performed in SAR11 bacteria, we tried to demonstrate the TMAO-importing function of TmoXWV₁₀₆₂ in a tmoW-deleted mutant of R. pomeroyi DSS-3. We constructed the mutant ΔtmoW_DSS–3 by deleting the majority of gene tmoW from the R. pomeroyi DSS-3 genome, and then complemented this mutant with tmoXWV₁₀₆₂ to generate the complemented strain ΔtmoW_DSS–3-tmoXWV₁₀₆₂ that contains the tmoXWV₁₀₆₂ cluster from strain HTCC1062. As shown in Figure 3, the ΔtmoW_DSS–3 mutant was unable to grow on TMAO, consistent with that previously reported (Lidbury et al., 2014). In contrast, the growth of the complemented strain ΔtmoW_DSS–3-tmoXWV₁₀₆₂ on TMAO was comparable to that of strain R. pomeroyi DSS-3, suggesting that the tmoXWV₁₀₆₂ cluster was involved in TMAO transport in strain ΔtmoW_DSS–3-tmoXWV₁₀₆₂. Considering that strain HTCC1062 can grow on TMAO (Figure 2A), this result indicates that tmoXWV₁₀₆₂ is most likely to encode a functional TMAO importer in strain HTCC1062.

The periplasmic substrate binding protein of an ABC transporter is usually responsible for the first-step recognition of substrate, and can bind a given ligand with high affinity (Albers et al., 1999; Chen C. L. et al., 2010). To characterize TmoX₁₀₆₂, the substrate binding protein of TmoXWV₁₀₆₂, the full-length tmoX₁₀₆₂ gene containing 934 nucleotides was amplified from the genome of strain HTCC1062 and was expressed in E. coli BL21(DE3) cells. To analyze the substrate specificity of TmoX₁₀₆₂, the binding affinities of the recombinant TmoX₁₀₆₂ toward TMAO, betaine, choline, TMA, DMA and carnitine were determined by ITC measurements. Among the tested substrates, TmoX₁₀₆₂ possessed a high binding affinity toward TMAO, with a K_d (dissociation constant) of 520 nM (Figure 4A), but presented little binding affinity toward betaine, carnitine, TMA or DMA (Figure 4 and Table 1). Compared to TmoX_DSS–3 of R. pomeroyi DSS-3, which exhibited a K_d of 1.6 μM toward TMAO (Li et al., 2015), TmoX₁₀₆₂ possessed a higher binding affinity toward TMAO. Considering the concentrations of TMAO range from nanomolar to low micromolar in marine environments (Gibb et al., 1999; Gibb and Hatton, 2004), the higher binding affinity of TmoX₁₀₆₂ toward TMAO would help strain HTCC1062 gain a competitive advantage over other bacteria at low TMAO concentrations. Surprisingly, the recombinant TmoX₁₀₆₂ also presented binding affinity toward choline, with a K_d of 2.5 μM (Figure 4B and Table 1). A similar phenomenon was also observed in TmoX_DSS–3 (Li et al., 2015). RT-qPCR results indicated that choline did not induce the transcription of tmoX₁₀₆₂ in strain HTCC1062 (Supplementary Figure 2), suggesting that the binding of recombinant TmoX₁₀₆₂ toward choline may not make physiological sense. Alternatively, strain HTCC1062 may utilize TmoXWV₁₀₆₂ as a multifunctional transporter to import both TMAO and choline, as this strain possesses a highly streamlined genome (Giovannoni et al., 2005; Sun et al., 2011; Noell and Giovannoni, 2019).

The tmoX gene is widespread in divergent marine bacteria, especially in SAR11 bacteria (Lidbury et al., 2014). The seawater temperatures are different at different depths and change with the seasons regularly, especially for surface seawaters (Malmstrom et al., 2010). Therefore, marine bacteria need to adapt different temperatures. To investigate the thermostability of TmoX₁₀₆₂, we measured the melting temperature (T_m) of TmoX₁₀₆₂. The T_m of TmoX₁₀₆₂ is 62.5°C (Figure 5A), which is higher than that of TmoX_DSS–3 (T_m = 54.5°C) (Li et al., 2015), suggesting that TmoX₁₀₆₂ has higher thermostability than TmoX_DSS–3. The binding affinities of TmoX₁₀₆₂ toward TMAO at different temperatures were also detected. TmoX₁₀₆₂ exhibited high binding affinities toward TMAO at 4°C (Figure 5B), 10°C (Figure 5C) and 25°C (Figure 5D), indicating that TmoXWV₁₀₆₂ should be able to import TMAO into cells of strain HTCC1062 efficiently at different temperatures. The nanomolar-level TMAO binding affinity, the high thermostability and the strong temperature adaptability of TmoX₁₀₆₂ may reflect the niche adaptation of HTCC1062 to the volatile marine environment, especially to the oligotrophic environment.

The TMAO binding mechanism of TmoX_DSS–3 has been proposed based on structural and mutational analyses (Li et al., 2015). In TmoX_DSS–3, the TMAO binding pocket is composed of Trp55, Trp102, Phe106, Glu131, Trp177, Phe220, and Trp222 (Figure 6A), among which Glu131 forms a hydrogen bond with the oxygen atom of TMAO, and four tryptophan residues (Trp55, Trp102, Trp177, and Trp222) form a rectangular aromatic box and interact with TMAO by cation-π interactions (Li et al., 2015; Figure 6A). The aromatic rings of two phenylalanine residues (Phe106 and Phe220) also participate in forming the hydrophobic cage to accommodate TMAO (Li et al., 2015; Figure 6A). TmoX₁₀₆₂ shares ∼51% sequence identity with TmoX_DSS–3, and the TMAO binding mechanism of TmoX₁₀₆₂ is still unclear.

To probe the TMAO binding mechanism of TmoX₁₀₆₂, we tried to co-crystallize TmoX₁₀₆₂ and TMAO and solve the crystal structure of TmoX₁₀₆₂. However, all the attempts failed. We then modeled the structure of TmoX₁₀₆₂ via Swiss-model² (Waterhouse et al., 2018), with the crystal structure of TmoX_DSS–3 (PDB code: 4XZ6) as the template. The overall structure of TmoX₁₀₆₂ is similar to that of TmoX_DSS–3 (Figure 6B), with a root mean square deviation (RMSD) between these two structures of 0.1 Å over 226 Cα atoms.

Structural analysis of the model of TmoX₁₀₆₂ showed that the binding pocket of TmoX₁₀₆₂ may be composed of Trp38, Trp85, Phe89, Glu114, Phe210, and Trp212 (Figure 6C), and Trp164, the corresponding residue of Trp177 in TmoX_DSS–3 (Figure 6A), may not participate in binding TMAO. The residue Glu114, corresponding to Glu131 in TmoX_DSS–3, may form a hydrogen bond with TMAO (Figure 6C). Mutation of Glu114 to alanine abolished the binding affinity of TmoX₁₀₆₂ toward TMAO (Figure 6D), indicating the important role of Glu114 in binding TMAO. The side chains of Trp38, Trp85 and Trp212 (corresponding to Trp55, Trp102 and Trp222 in TmoX_DSS–3, respectively) form an aromatic box (Figure 6C). Together with the side chains of Phe89 and Phe210 (corresponding to Phe106 and Phe220 in TmoX_DSS–3), this box forms a hydrophobic cage to accommodate the quaternary amine of TMAO (Figure 6C). Mutations of Trp85, Phe89 and Trp212 to alanine severely deceased the binding affinity of TmoX₁₀₆₂ toward TMAO, suggesting the important roles of these three residues in substrate binding (Figure 6D). However, mutants W38A and F210A still maintained a relatively high TMAO binding affinity (Figure 6D), indicating that these two residues may not participate in TMAO binding, or the other residues of TmoX₁₀₆₂ may compensate the function of Trp38 and Phe210. In TmoX_DSS–3, mutations of the corresponding residues composing the TMAO binding pocket all decreased its TMAO binding affinity to a large extent (Li et al., 2015). Therefore, our biochemical results suggested that several residues of TmoX₁₀₆₂ participating in TMAO binding may be different from those of TmoX_DSS–3, although sequence analysis showed that the residues comprising the binding pocket of TmoX are all highly conserved in the MRC and the SAR11 clade (Li et al., 2015). CD spectroscopy assays showed that the secondary structures of the mutants are similar to that of WT TmoX₁₀₆₂ (Figure 6E), indicating that the decrease in the binding affinities of the mutants is a result of residue replacement rather than structural changes.